You’ve obviously noticed that no one in your home has the same Netflix homepage — the shows we’re recommended align directly with our preferences and they change as our tastes do. You’ve probably also heard people say how much of a time-suck TikTok is. That they didn’t even realize as hours pass by while they scroll through the app.
That’s machine learning in action.
A subset of artificial intelligence, ML constantly uses past data and results to progressively make better decisions in the future. It’s ushering in a new wave of efficiency: It doesn’t require millions of lines of code as it goes to work on multiple scenarios at incredible speed.
What AutoML Means for Business
Machine learning multiplies value for organizations, from cutting costs to driving growth. When using ML, however, it’s not uncommon for companies to face certain problems.
In particular, Statista found that the top two ML challenges right now are scaling up (43%) as well as versioning and reproducibility in ML models (41%). The lack of data scientists proficient in artificial intelligence and ML, and the need to generate insights faster, fuel these issues.
This is where AutoML comes in.
AutoML automates the process of deploying ML algorithms, from data processing to hyper-parameter tuning. At its core, AutoML is a set of algorithms that automate other machine learning algorithms.
This can do at least four things for businesses and industries:
- Reduce the time it would take to deploy ML processes
- Improve accuracy of ML models
- Boost productivity by focusing on ML optimization
- Allow more companies, especially those with limited expertise, to use ML
Where Machine Learning is Headed in 2020 and Beyond
AutoML is one of the most popular and easily available breakthroughs in the area of machine learning. From 2020 onward, the vision is that more companies will be able to do things that only large organizations can do right now. Such as:
1. Powering mobility, transportation, and logistics
Solving travel behavior, transportation, and even people’s movement issues is possible with machine learning techniques. MIT’s JTL Urban Mobility Lab moves beyond “the traditional approach of using discrete choice models” to a “deep neural network to predict individual trip-making decisions and to detect changes in travel patterns.”
Concrete applications include developing dynamic pricing for ride-hailing services, rebalancing fleets, optimizing routes, and detecting anomalies.
US-based Convoy, a logistics company, uses machine learning to better match shippers and truckers on its online marketplace. Its ML models process millions of shipping jobs together with trucker availability. A similar ML perspective is also used by Lyft and Grab.
Another great example is the Dubai-based company, Aramex. They use machine learning to implement a chatbot and create an intelligent address prediction model. As a result, it’s able to improve communication and delivery windows and thus, the customer experience.
What’s in it for retail? A look at machine learning for eCommerce
In retail, data generated from mobility sensors such as GPS, smart cards, WiFi, and digital displays can help brands improve operations and customer experiences. With ML, retailers can scale up the delivery of hyper-relevant offers, loyalty-building moments, and omnichannel engagement.
2. Risk management: from cybersecurity to revenue loss
Machine learning is increasingly being used in business applications such as finance and healthcare. With ML, organizations can gain insights that will help optimize processes faster, from detecting fraudulent activities to achieving better health outcomes. PayPal, for instance, uses AI and ML in various applications, like customer service and autonomous automation.
However, the use of machine learning in itself involves risks. Its very nature relies on data and predictions. Questions arise about the stability of data, the applicability (or not) across population segments, consumer privacy, and more.
Because machine learning as a field and business tool is still young, this is a development worth considering when planning to use it.
What’s in it for retail?
Machine learning can help retailers better prepare for the future of sudden shifts in supply and demand and overall uncertainty. A study found that 73% of retailers believe AI and ML can be valuable in demand forecasting — this is particularly pertinent as COVID-19 shifts spending habits and consumer behavior.
According to Brian Kilcourse, Managing Partner at RSR, “With such unpredictability, the ability to be agile and model potential outcomes becomes even more important. Retailers need AI-enabled predictive models for things such as labor and transportation costs across the supply chain or finding optimal DC-to-customer locations to lower costs while still satisfying rapidly changing customer needs.”
3. Regulating technology misuse
While new technology brings with it enormous benefits, its misuse is increasingly becoming a concern for businesses, consumers, and governments. In AI, deepfakes have the potential for distributing false content which then can influence people’s opinions.
Adobe, together with scientists from UC Berkeley, plans to use machine learning to automatically detect fake photos, in particular, manipulated images of faces. This is in addition to its tool that detects edited media such as videos.
Another example is the use of obtrusive ads on apps. Besides poor user experience, these disruptive and “out-of-context” ads, as Google calls them, also waste advertisers’ budgets. Using ML, Google automatically detects poorly placed ads and removes the apps that were displaying them.
What’s in it for retail?
In retail, technology misuse isn’t new. Recently, Bloomberg released an article about how voice-powered devices of tech giants such as Apple and Google are harvesting and analyzing people’s conversations.
Customer data has become a competitive asset for retailers. A potential solution for preventing the abuse of data is deploying privacy-preserving machine learning models.
One approach is called differential privacy which designs ML algorithms that extract insights from data while concealing information. Recently, IBM released a differential privacy library that works with ML and only a single line of code. Other companies researching in this area include Apple, Google, and Uber.
Moving forward, as research on privacy-preserving ML models become more accessible to the public, organizations of all sizes can adapt. As a result, businesses can unlock more capabilities and utility from data.
4. Personalizing every customer touchpoint
Speaking of machine learning for eCommerce, its most exciting and powerful application will be on personalization. With ML, retailers can turn data into insights at scale. This will support the selection of the best experience for individual customers.
According to McKinsey & Co., having the right personalization in place increases revenue by 5% to 15% and marketing-spend efficiency by 10% to 30%. Moreover, three major shifts will make personalization more personal: Digitizing physical spaces, scaling empathy, and using ecosystems to personalize the journey.
Target, for instance, already invests heavily in machine learning for eCommerce to improve product recommendations and optimize advertising spend. Overstock has real-time personalization on its website to ensure offers are relevant to each customer.
Using machine learning for eCommerce personalization is a growing trend. But consumers also have questions over privacy, especially when brands do things that are perceived as invasive.
Accenture found that “83% of consumers are willing to share their data to enable a personalized experience as long as businesses are transparent about how they are going to use it and that customers have control over it.”
In sum, machine learning is a subset of AI that can be applied in different areas of business. With AutoML, more companies can leverage ML capabilities that only bigger organizations have access to today.
Across industries, it’s important to balance innovation with security and privacy. Doing so ensures better customer experiences and helps build stronger brand loyalty based not only on convenience but also on trust.